Overview

Designing an AI-Powered Chatbot for Patients

Crafting a patient-focused AI chatbot for healthcare provider websites

MY ROLE
UI Design
Research

TOOL
Figma

DURATION
Dec 2023
(1 Month)

Team
Designer (Me)
1 Product Manager
2 Developers

Introduction

At Klarity Health, I designed a patient-focused AI chatbot as an add-on service for healthcare provider websites. This chatbot helps patients quickly find information on complex or detailed doctor sites, enhancing their browsing experience. The project is set to launch in 2024.

Problem

Healthcare providers face challenges in managing patient inquiries efficiently, leading to increased response times and patient dissatisfaction.

Project goal

Business goal

Provide 24/7 support and information to potential patients, improving communication and satisfaction.

Drive sales of Klarity's B2B product by empowering healthcare providers with online practices and attracting potential patients.

Design process

Discover

Surveys

I designed a 12-question survey and collected feedback from 30 target users to understand their needs when visiting provider websites.

Key insights

• Provider's background and appointment availability are important to users.
• Users prefer bot communication over text-only interactions.

Competitive analysis

Given the time limitations, I conducted a competitive analysis to gather insights from existing chatbot solutions in various industries.

Define

How does the AI chatbot works?

The AI chatbot uses ChatGPT for natural language processing and machine learning. Collaborating with the PM, we developed training questions for the chatbot. The backend engineer then trained the chatbot to handle user queries effectively.

User flow

Bot persona

I transformed insights from competitive analysis and user surveys into design solutions, refining the AI chatbot's usability, functionality, and visual design.

Design

Ⅰ. Onboarding

The chatbot ensures a smooth onboarding experience with two options to initiate a conversation
1. Pre-defined common replies for faster 
information exchange.
2. Allows users to type specific queries or concerns.

Ⅱ. Service requests

The chatbot handles service requests with two options, and option 1 is chosen for its user convenience and immediate booking capability.
✅ Users see available times directly in the chat and click on a time to book immediately.
❌ Users click the booking link to see all available times.

Ⅲ. Offline message

Users can leave messages for providers for direct communication, such as specific questions about medication, ensuring effective communication and addressing individual needs.

Ⅳ. Error-handling

Effective error handling is crucial. Potential issues include: random/unrelated input, requests for untrained tasks, complex expressions, and small talk.
1. Analyzes relevant content if input is outside training.
2. Reminds users to focus on relevant questions and offers quick replies for long or error-filled inputs.

Ⅴ. Evaluate services

Collect user feedback and satisfaction ratings. Upon closing a chat session, users are prompted to rate their experience.

Internal testing

Usability testing feedback

After the design phase, we conducted internal usability testing with 10 participants. The main task was to complete an "Appointment with a provider" using the chatbot.

1. No feedback flagging

Users can't flag incorrect or unsatisfactory answers.

2. Inconvenient location access

Users must copy the address to a map app for clear viewing.

3. Limited evaluation options

Users can only select a satisfaction level without providing additional comments.

Iterations

Ⅰ. Improved location display

Replaced address with an interactive map for better accessibility.

“Is there an easier way to access the location information of the doctor's office?"
- Team member

Ⅱ. Feedback text input

Added a text input for users to provide overall feedback during evaluation.

“Where can I write overall feedback?"
- Team member- Team member

Final deliverable

After iterative design and rigorous internal testing, the chatbot is ready for healthcare providers' websites. The official launch is scheduled for 2024, benefiting both providers and patients.

Chatbot design system

Success metric

How would we measure success?

To measure the chatbot's success, Klarity will track:
• Conversion rate
• User engagement
The chatbot will go live in 2024, alongside Klarity's new provider site platform.

Reflections

What I learned

This project taught me the importance of strong design rationale and stakeholder communication. Taking initiative, asking insightful questions, and owning projects led to fresh perspectives and better solutions. Overall, the experience underscored the value of proactive problem-solving and continuous learning.